Improving the quantiﬁcation of climate change hazards by hydrological models: A simple approach for mimicking the impact of active vegetation on potential evapotranspiration
- 1Institute of Physical Geography, Goethe University Frankfurt, Frankfurt am Main, Germany
- 2Senckenberg Biodiversity and Climate Research Centre (SBiK-F), Frankfurt am Main, Germany
Abstract. Almost no hydrological model takes into account that changes in evapotranspiration are affected by how the vegetation responds to changing CO2 and climate. This severely limits their ability to quantify the impact of climate change on evapotranspiration and thus water resources. We developed a simple approach for mimicking, in hydrological models, the impact of active vegetation on potential evapotranspiration (PET) under climate change. This approach can be applied for climate change impact studies by hydrological models that compute PET as a function of net radiation and temperature only, i.e., using the Priestley-Taylor (PT) equation. Our approach is based on the work of Milly and Dunne (2016) (MD), which compared the change of non-water-stressed actual evapotranspiration (NWSAET) as computed by an ensemble of global climate models (GCM) with various methods for computing PET change. MD proposed to estimate the impact of climate change on PET as a function of only the change in net energy input at the land surface, i.e., to neglect changes in other climate variables. The new mimicking approach (PT-MA) for application in hydrological models retains the impact of temperature on daily to interannual as well as spatial PET variations but removes the impact of the long-term temperature trend on PET such that long-term changes in future PET are driven by changes in net radiation only. We implemented the PT-MA approach in the global hydrological model WaterGAP 2.2d and computed daily time series of PET between 1901 and 2099 using the bias-adjusted output of four GCMs for RCP8.5. With PT-MA, increases of GCM-derived NWSAET between the end of the 20th and the end of the 21st century are simulated well by WaterGAP, while severely overestimated with the standard PT. Application of the mimicking approach in WaterGAP results in smaller future decreases or larger future increases in renewable water resources (RWR) as compared to neglecting active vegetation, except in a small number of grid cells where increased inflow from upstream due to increased upstream runoff leads to enhanced evapotranspiration from surface water bodies or irrigated fields. On about 20 % of the global land area, the mimicking approach leads to an increase of RWR that is more than 20 % higher than when neglecting the active vegetation, while on more than 10 % of the global land area, the projected RWR decrease is lowered by more than 20 %. We recommend applying the mimicking approach to assess climate change hazards by any hydrological model that does not include stomatal conductance when calculating PET.
Thedini Asali Peiris and Petra Döll
Status: final response (author comments only)
RC1: 'Comment on hess-2022-230', Anonymous Referee #1, 29 Aug 2022
- AC1: 'Reply for RC1', Ardhige Thedini Asali Peiris, 22 Oct 2022
RC2: 'Comment on hess-2022-230', Anonymous Referee #2, 27 Sep 2022
- AC2: 'Reply for RC2', Ardhige Thedini Asali Peiris, 22 Oct 2022
Thedini Asali Peiris and Petra Döll
WaterGAP2.2d model derived Potential evapotranspiration and Renewable water resources variables under climate change, with and without vegetation response mimicking approach. https://doi.org/10.5281/zenodo.6593136
Thedini Asali Peiris and Petra Döll
Viewed (geographical distribution)
In this study, the authors document an approach that aims to represent the influences of vegetation responses to rising CO2 and climate changes on potential evapotranspiration (PET) in hydrological models. The approach is a modified version of Priestley-Taylor (PT), which represents PET as a function of only the change in net energy input and temperature, but removing the long-term temperature trend (PT-MA). The approach is implemented in the WaterGAP model, which, when driven by output from historical and future (e.g., RCP8.5) global climate model simulations, is shown to capture the PET in non-water-stressed regions well compared to PT. Overall, the PT-MA method leads to a smaller increase in PET than PT, so there is a relatively smaller decrease (or larger increase) in water resources in the future.
Overall, the paper is well written, and the method presented has utility. However, there are some flaws in the analysis/interpretation and the framing of the method as "representing the effects of vegetation" is misleading. It would be more accurate to say that PT-MA accounts for processes that might oppose the influences of long-term warming on increasing PET, since it does not actually "mimic" the direct effects of vegetation. I think the paper could be publishable after a significant rewriting and reframing of the results, as well as some additional analysis (e.g., statistical significance testing). Specific suggestions are outlined in the major and minor comments below.
Line 10: It is not clear in the abstract how the approach attempts to capture the effects of "active vegetation". I suggest adding a sentence or two that explains/justifies this connection more directly.
Line 110: Why is the approach only validated against 3 (or 4) GCMs? Why not use all 16 models that MD used? What criteria were used to choose the GCMs that were used in this study?
Line 124: This paragraph describes how the different components of evapotranspiration are calculated, as dependent on PET. Is the PT-MA method for calculating PET applied to all components or just some? Further down (line 175) it is stated that modified T is not used for open water, but what about regions where there is little vegetation (i.e., leaf area index is low), why use a version of PET adjusted for vegetated conditions in those regions? Also, vegetation effects would mostly have an influence during the growing season, so would it be more appropriate to use the original approach during winter?
Line 126: Since canopy evaporation is calculated as a function of PET and leaf area index, is leaf area index from the GCM simulations also used? Leaf area increases due to rising CO2 and will influence both canopy evaporation and transpiration.
Line 170: Given that vegetation effects will be most important during the growing season, why not remove the long-term monthly (or seasonal) trend? For example, if the summer is warming more quickly than the winter, would it be more appropriate to remove the summer trend rather than the annual trend?
Line 235: How is it possible that actual ET is higher than net radiation (i.e., HadGEM2-ES)? What is the energy balance of net radiation, latent heating, sensible heating, and ground heat flux?
Line 244: A fixed value of 0.8 would likely not apply in all regions (as the authors discussion in section 4.2), so it makes sense that it would not match the PT-MA results. It could be that it is less than or more than 80% of net radiation in some regions. If you calculated the net radiation directly from the GCMs to compare to WGHM, you could determine if it is the radiation that is different or if it is the "scaling factor", which would help the discussion here.
Figure 2/3: Why is there no difference between PT vs PT-MA (and T vs modified T) before 2000? With a reference period of 1981-2000, it would be assumed that actual temperature T would be lower than the modified temperature for the early 1900s, since a long-term warming trend would be present from 1900 to 2000.
Line 260: Why not include all the models in Figure 3? Since there are only 4 GCMs, it seems somewhat random to choose 2 to put in the main manuscript and 2 others for the appendix.
Line 263: Why is it intended that there is no difference before 2001? Was there no long-term temperature trend in those locations prior to 2001? Why do you not remove the temperature trend for the entire time series?
Line 272: Again, why are these 2 GCMs chosen over the other 2? Or why not show the multi-model mean?
Line 291: I suggest removing "and open water bodies", since there wouldn't be any limit on water availability in open water bodies.
Line 296: With the color bar used in Figure 5 and no statistical significance testing, it is not possible to determine if the changes in the western US are meaningful. I suggest adding a color (white) centered on zero to indicate regions with no change. And adding some statistical significance testing to these figures.
Line 303: As mentioned in the main comment above, the DC metric may appear to show large differences where the climate change signals are small. I suggest adding some indication in this figure for where the climate change signal is significant.
Figure 6: Again, with the color bar it is not possible to determine if the differences are meaningful for panels a and b. Very small values (e.g., -0.00001 would appear as yellow and 0.00001 would appear as light green) could be from rounding error. I suggest adding a color (white) that is centered on zero (e.g., -0.5 to 0.5), and adding significance testing.